A Study on of Music Features Derived from Audio Recordings Examples

Home , Alex Band, The Calling, What the Hell, Wherever You Will Go

ARCHIVES OF ACOUSTICS Copyright c 2018 by PAN – IPPT Vol. 43, No. 3, pp. 505–516 (2018) DOI: 10.24425/123922

A Study on of Music Features Derived from Audio Recordings Examples – a Quantitative Analysis

Aleksandra DOROCHOWICZ(1), Bożena KOSTEK(2)

(1) Multimedia Systems Department Faculty of Electronics, Telecommunications and Informatics Gdansk University of Technology Narutowicza 11/12, 80-233 Gdańsk, Poland; e-mail: [email protected]

(2) Audio Acoustics Laboratory Faculty of Electronics, Telecommunications and Informatics Gdansk University of Technology Narutowicza 11/12, 80-233 Gdańsk, Poland; e-mail: [email protected]

(received June 19, 2017; accepted March 21, 2018)

The paper presents a comparative study of music features derived from audio recordings, i.e. the same music pieces but representing diﬀerent music genres, excerpts performed by diﬀerent musicians, and songs performed by a musician, whose style evolved over time. Firstly, the origin and the background of the division of music genres were shortly presented. Then, several objective parameters of an audio signal were recalled that have an easy interpretation in the context of perceptual relevance. Within the study parameter values were extracted from music excerpts, gathered and compared to determine to what extent they are similar within the songs of the same performer or samples representing the same piece. Keywords: music genres; audio parametrization; music features.

1. Introduction account a number of feature types at the same time. Music genres are also divided into smaller sub-groups There are many formal systems describing differ- (punk rock, new wave, post grunge) and they can be ences between music pieces, which allow for defining mixed (symphonic metal, pop rock). its assignment to a category or type of music. Many Throughout the history, many types of classifica- features influence the final categorization of the song, tions were created. They were based on music styles, some of them can be described or measured as is forms and genres. Over time, the topologies become done within the Automatic music genre classification more and more complex and divided. Moreover, the (AMGC) area (Bergstra et al., 2006; Bhalke, 2017; pieces could belong to more than one category, so it is Burred, Lerch, 2014; Hoffmann, Kostek, 2015; necessary to identify their definitions and significance. Kalliris et al., 2016; Kotsakis et al., 2012; Kostek, These include musical form, style, genre and descrip- 2005; Ntalampiras, 2013; Schedl et al., 2014; Silla tive features such as meter, tempo, structure or origin. et al., 2007; Sturm, 2014; Tzanetakis et al., 2002; The definition of the music form is based on the vocal Rosner et al., 2014; Rosner, Kostek, 2018). One and/or music instruments used, texture type or num- of the ways the songs can be described is by using ber of passages it consists of (one-passaged, cyclic). their origin (J-rock, Brit pop), time when they are Moreover, the Small Music Encyclopedia describes mu- composed or performed, performance (Pluta et al., sic forms as the typical for a specific group of music 2017), mood of music (Barthet et al., 2017; Plewa, piece schemes, designated by the analysis based on the Kostek, 2015), instruments used (symphonic, acous- specific pieces. tic, rock), music techniques (riff, rap) or function that At the beginning of 17th century, the concept of music plays (film score, religious or orchestral music) the style in the music theory changed its meaning (Benward, 2003). Frequently, to be able to conduct (Pascall, 2001; Seidel, Leisinger, 1998). Division a proper analysis of a piece, one may need to take into based on antique tradition: religious, chamber and 506 Archives of Acoustics – Volume 43, Number 3, 2018 scenic (theatrical), within subtypes were proposed by et al., 2006; Tzanetakis et al., 2002; Holzapfel, Marc Scacchi in 1649 (Palisca, 1998). Athanasius Stylianou, 2008; Kostek et al., 2011; Kostek, Kircher proposed to expand it into church, madrigal Kaczmarek, 2013; Ntalampiras, 2013), where the and theatrical style, noting that it is not the place user may choose a song belonging to the particular of performance that specifies the style, but the ef- music genre (Tzanetakis et al., 2002). fects triggered. His proposal also predicts styles dif- The aim of the study is to determine to what extent fusions (Helman, 2016). Although the definition of music feature values and characteristics of music ex- the music style has changed through the centuries, cerpts are similar for the performer or for a music piece. according to the Small Music Encyclopedia, the mu- Tracks gathered for the purpose of this study are de- sic style is the term describing the common features scribed by their parametric representation, i.e. time of the compositional technique typical for the specific and spectral parameters, such as: RMS (Root-Mean- piece, author, for the national music, historical period Square) energy, zero-crossing rate, spectral centroid, (Dziębowska, 1998). spectral skewness, spectral kurtosis, spectral flatness, A music genre identifies some pieces of music as entropy of spectrum, brightness and roll-off, all ex- belonging to a shared tradition or set of conventions. tracted by the Matlab MIRtoolbox1.6.2 (MIRtoolbox). Music genres are created based on many types of di- At the same time several descriptive features such as visions: age of the recipients (e.g. adult contemporary, music genre, time of origin, country are identified. Mu- teen pop, music for children), artists’ origin (e.g. Amer- sic excerpts used in the experiment represent various icana, Afro-Cuban, Brit pop, Italo disco, K-pop, etc.), music genres, assigned according to the artist’s ambi- time of the origin (e.g. baroque, classical, contem- tions, origin, time of origin, they include various pieces porary, etc.), artists’/music ambitions (classical mu- of the same artist, and the same songs performed by sic, ambient, country, etc.), ideology (e.g. rock, hip- a number of various artists. hop, rap, metal, blues, New Age, etc.), instrumenta- tion and treatment of musical instruments within the given genre (e.g. jazz, country, folk, electronic, blues, 2. Experiments rock, etc.). 2.1. Building a database The American music magazines Alternative Press (Alternative Press, 2016) and Rock Sound (Rocksound, In the study, 202 music excerpts, 0.5–1 minute long, 2016) at the beginning of the 2016 drew attention to were collected. They represent various music genres, the problem of lack of the unequivocal music genres i.e.: rock, pop-punk, new wave, glam metal, punk rock, definition and divisions in the research study entitled Brit pop, pop, soft rock, blues, musical, rock and roll, “What is punk?”. It was carried out by the Converse psychedelic rock, soul, art rock, heavy metal, emo, company in associate with the Polygraph’s analyst post grunge, cabaret, pop rock, different time of ori- (Daniels, 2016). Their bases were playlists tagged as gin (from 1943 to 2016), different countries and instru- punk at Spotify and YouTube’s playlists. Over half of ments used. Also, the research includes samples of one them (51%) contain such music bands as: Green Day, song performed by various artists and various songs Blink-182 (50%), The Offspring (44%), Sum 41 (39%), performed by one artist. Rise Against (38%), Fall Out Boy (38%), My Chemi- Moreover, in our previous study we have performed cal Romance (35%), Bad Religion (30%), Nofx (28%), listening tests to eventually correlate the subjective All Time Low (28%) and A Day to Remember (27%). session outcome with the artificial intelligence algo- All the bands belong to genres closely related to punk rithms results (Dorochowicz et al., 2017). The task (pop-punk, emo, post-hardcore, metalcore), although of the test participant was to assign a music excerpt to they are not strictly punk. This research shows that it the specific music genre. The same goal was expected is not obvious to which category the performer or the to be achieved by the machine learning approach. Even song belongs, which makes defining it in some cases though the listening test results were not fully homo- almost impossible. geneous due to the individual characteristics of mu- It should also be noted that within the Music In- sic as well as the ambiguity of some tracks, causing formation Retrieval area there exists a considerable that people attributed a given track to different mu- disagreement among the researchers over music genre sic genres, still the results were more than encourag- classification, as assigned by a human or by a ma- ing (Dorochowicz et al., 2017). We have obtained chine learning algorithm, i.e. whether such an assign- a good agreement between subjective and “objective” ment is practical (Silla et al., 2017; Tekman, Hor- approaches, the latter one based on machine learning. tacsu, 2002). At the same time there are a lot of ex- This helps to determine which music excerpts can be amples of research studies, as well as commercial appli- considered as uniquely attributed to a particular genre, cations (e.g. music social networking, music catalogu- checked both by subjective tests and machine learning tools for applications, etc.) related to automatic ing approach. Unambiguous cases obtained in that way genre recognition and classification (e.g. Bergsta were parametrized and further analysed. A. Dorochowicz, B. Kostek – A Study on of Music Features Derived from Audio Recordings Examples. . . 507

2.2. Parametrization • Spectral centroid – returns the first moment (mean), which is the geometric center (centroid) As already mentioned the analysis performed is (Fig. 2). It is interpreted as a measure of Bright- based on a quantitative comparison of parametric rep- ness (center of gravity; see Fig. 3): resentation of music excerpts. To parametrize music Z excerpts several features have been chosen, as defined µ = xf(x) dx. (2) by the Matlab MIRtoolbox 1.6.2 (MIRtoolbox). The criterion for choosing them was a potential physical • Spectral skewness – is the third central moment, interpretation and perceptual relevance of a descrip- showing the asymmetry of the distribution around tor. They were as follows: its mean. When the value is positive, the distribution has a longer tail to the right, the negative • RMS energy – Root-Mean-Square Energy; a pa- value means the opposite. When the value equals rameter showing the global energy of the signal: zero, the distribution is symmetrical (Fig. 4): v Z u n r 2 2 2 3 u 1 X 2 x1 + x2 + ... + xn µ3 = (x − µ1) f(x) dx. (3) xrms = t x = . (1) n i n i=1 • Spectral kurtosis – is defined as the fourth stan- • Zero-crossing rate – counts how many times the dardized moment minus 3 (to correct the kurtosis signal crosses the X-axis (Fig. 1), and it is related of the normal distribution equal to zero). Its inter- to the noisiness of the signal. pretation is associated with flatness of the spectral distribution around its mean. Examples of the Also, several spectral distribution-based parame- kurtosis figures are shown in Fig. 5. ters were used that can be described by statistical mo- • Spectral flatness – defined as the ratio of the ments: centroid, spread, skewness, kurtosis, flatness, as geometric and arithmetic means of the coefficients well as entropy:

Fig. 1. Signal crossing X-axis (MIRtoolbox).

Fig. 2. Centroid (MIRtoolbox).

Fig. 3. Brightness (MIRtoolbox). 508 Archives of Acoustics – Volume 43, Number 3, 2018

of the power density spectrum in every spectral 3. Analyses bands (b) of the width of 1/4 octave (Eq. (4)). This feature is also called a tonality coeﬃcient Examples of the analysis results obtained are shown (Dubnov, 2004) below. They concern a quantitative analysis of parameter values extracted from the music excerpts. The music excerpts analysed are divided into the following  1  "N/2 # N/2 groups: Q j 2πk  P e N  • Artists’ songs for whom parameters val-  k=1  SFM = 10 log  , (4) ues are similar – a thorough analysis of songs 10  N/2   1 P j 2πk  that are potentially similar in perception was per-  P e N  N/2 k=1 formed, and among the music excerpts gathered such artists as listed below are to be mentioned:

j 2πk where P e N is the PSD calculated on the ba- – Green Day (see Fig. 7) – punk rock, songs: sis of the N-point DFT. Basket Case, Blitzkrieg Bop (cover of Ra- mones – punk rock), Like a Rolling Stone • Entropy of spectrum – a measure that describes (cover of Bob Dylan – country), Tired of spectrum uniformity. Equation (5) returns the rel- Waiting (cover of The Kinks – rock), ative Shannon entropy: n – Elaine Paige (musical singer, songs: Don’t X Cry For Me Argentina (cover of Madonna – H(X) = − p (xi) logb p (xi) . (5) i=1 pop), I Dreamed a Dream, Memory), • Roll-off – estimates the amount of high frequency – The Blues Brothers – rock’n’roll, Every- in the signal by finding the frequency below which body Needs Somebody to Love, Jailhouse Rock 85% of the magnitude distribution is concentrated (cover of Elvis Presley – rock’n’roll), Sweet (Fig. 6). Home Chicago; • Artists’ songs for whom parameter values differ, examples of them are listed below: – Sarah Brightman (see Fig. 8) – musical singer, songs: Don’t Cry For Me Argentina (cover of Madonna – pop), Memory (cover Fig. 4. The negative and positive value of the skewness of Elaine Paige – musical singer), My Heart (MIRtoolbox). Will Go On (cover of Celine Dion – pop), Time To Say Goodbye, – The Police – new wave, songs: Roxanne, Can’t Stand Losing You, Spirits in the Mate- rial World, – The Calling – post grunge, songs: London Fig. 5. Different values of the kurtosis (MIRtoolbox). Calling (cover of The Clash – punk rock),

Fig. 6. Roll-oﬀ parameter illustration (MIRtoolbox). A. Dorochowicz, B. Kostek – A Study on of Music Features Derived from Audio Recordings Examples. . . 509

Fig. 7. Parameter values (normalized) similar for music excerpts of a singer (Green Day).

Fig. 8. Parameter values (normalized) diﬀerent for music excerpts of the same singer (Sarah Brightman).

Our Lives, Things Will Go My Way, When – Helena, I’m Not Okay (I Promise), Na Na It All Falls Down, Wherever You Will Go; Na (Na Na Na Na Na Na Na Na Na), Song 2 (cover of Blur – Brit pop), Welcome to the • Artists, whose cover and their own songs Black Parade, have diﬀerent characteristics, such as: – All Time Low – pop punk, songs: Bad Rep- – Avril Lavigne – pop punk, songs: All The utation (cover of Joan Jett – punk rock), Small Things (cover of Blink-182 – punk Blitzkrieg Bop (cover of Ramones – punk rock), Bad Reputation (cover of Joan Jett – rock), For Baltimore, Kids in the Dark, Miss- punk rock), Basket Case (cover of Green Day ing You, Poppin’ Champagne, Runaways, – punk rock), Girlfriend, Here’s To Never Should I Stay or Should I Go (cover of The Growing Up, How You Remind Me (cover Clash – punk rock); of Nickelback – post grunge), Knockin’ On Heaven’s Door (cover of Bob Dylan – coun- • Artists, who have an original song diﬀerent try), Sk8er Boi, Stop Standing There, The from others contained in their discography, Best Damn Thing, What The Hell; shown in such as e.g.: Table 1, – Peggy Lee – jazz, songs: He’s a Tramp, I’m – My Chemical Romance – emo, songs: Deso- a Woman, It’s a Good Day, Why Don’t You lation Row (cover of Bob Dylan – country), Do Right (shown in Fig. 9), 510 Archives of Acoustics – Volume 43, Number 3, 2018

Table 1. Covers diﬀerent than artist’s own songs (Avril Lavigne) (covers are marked in grey colour).

Zero- RMS Spectral Spectral Spectral Spectral Entropy Song crossing Brightness Rolloﬀ energy centroid skewness kurtosis ﬂatness of spectrum rate Girlfriend 0.62 0.55 0.60 0.44 0.24 0.17 0.95 0.84 0.57 Here’s To Never 0.81 0.59 0.81 0.39 0.20 0.61 0.97 0.86 0.82 Growing Up Sk8er Boi 0.64 0.62 0.66 0.49 0.28 0.23 0.97 0.88 0.61 Stop Standing There 0.69 0.57 0.68 0.47 0.25 0.41 0.96 0.85 0.70 The Best Damn Thing 0.83 0.66 0.60 0.45 0.25 0.22 0.96 0.84 0.59 What the Hell 0.79 0.70 0.70 0.52 0.28 0.42 0.96 0.87 0.69 All the Small Things 0.51 0.67 0.78 0.27 0.17 0.21 0.97 0.94 0.71 Bad Reputation 0.48 0.66 0.83 0.28 0.16 0.27 0.97 0.94 0.80 Basket Case 0.43 0.76 0.81 0.30 0.17 0.31 0.98 0.92 0.76 How You Remind Me 0.47 0.52 0.64 0.41 0.21 0.21 0.95 0.82 0.71 Knockin’ 0.38 0.71 0.95 0.21 0.13 0.31 0.98 0.92 0.91 on Heaven’s Door Standard Deviation 0.05 0.09 0.11 0.08 0.03 0.05 0.01 0.05 0.08

Fig. 9. One diﬀerent song of Peggy Lee.

– or Madonna – pop: Don’t Cry For Me Ar- – or Meat Loaf (see Fig. 10) – art rock, songs: gentina, Frozen, Material Girl; I’d Do Anything For Love (1993, Bat Out of Hell II), Rock And Roll Dreams Come • Artists, whose style evolved over time, Through (1993, Bat Out of Hell II), It’s All among them there are: Coming Back To Me Now (2006, Bat Out of Hell III), Seize The Night (2006, Bat Out of – Offspring – punk rock, songs: Blitzkrieg Bop Hell III), What About Love (2006, Bat Out (cover of Ramones – punk rock) (1984), of Hell III). Pretty Fly (For a White Guy) (1998), The Kids Aren’t Alright (1998), You’re Gonna Go Although not all the parameters of music samples Far, Kid (2008), of one group are very close to each other or may even A. Dorochowicz, B. Kostek – A Study on of Music Features Derived from Audio Recordings Examples. . . 511 manifest different tendency, their similarity/dissimi- For this case a Test of Significance for Two Un- larity can be judged on the basis of the standard devia- known Means and Unknown Standard Deviations (ac- tion. In Fig. 7, one can observe some parameter values cording to Eq. (6) was performed (Lane, 2017), which that are spread out across the scale. For example, the showed that for both significance levels, i.e. 5% and zero-crossing rate has values from 0.34 up to 0.59, on 1% (Soper, 2017), for three parameters: RMS-energy, the other hand, values of spectral kurtosis parameter spectral skewness and spectral kurtosis the differences are within the range of 0.18 to 0.22. The biggest value are statistically significant. of the standard deviation is for the zero crossing pa- s rameter and is equal to 0.11, whereas the smallest one 2 2 s1 s2 is nearly neglectful (i.e. 0.0095 for the entropy). t = + , (6) n1 n2 Parameters of music samples in Fig. 8 have in most cases different values, apart from entropy. However, 2 2 where s1, s2 – variances of the two groups, n1, n2 – there is no visible pattern of what values do the pa- size of the groups. rameters take. Most of them differ from each other. Parameters of the first song contained in the chart The largest value of the standard deviation is 0.33 for in Fig. 9 differ as to their values comparing to other spectral skewness. songs, this especially concerns Spectral skewness, Spec- In the case of Table 1 one may discern two groups tral kurtosis (both have higher values with the stan- of results, namely: samples of the songs owned by the dard deviation approx. equals 0.24) and to some extent artist (the first six songs), and the second one contains – Brightness (lower value). This song was written for covers performed by this artist. Samples of the songs different purpose, i.e. it comes from a movie, that is originally performed by the artist have higher RMS why it affected the artist’s performance. energy, lower spectral centroid, higher spectral skew- In the case of differences during the period of time ness and kurtosis. Contrarily, covers have in most cases (Fig. 10), two groups of the results coming from two higher zero-crossing range and lower spectral flatness. albums: one from 1993 and the other from 2006 are It is an interesting observation as one may hypothe- presented for this particular artist (Meat Loaf). It can size that the artist has tried to imitate performance be observed that for the newer songs values are much of the covers or did not use the whole potential when higher for Spectral flatness, contrarily the older songs performing covers. have higher Spectral centroid, skewness and kurtosis.

Fig. 10. Diﬀerences over the period of time of an artist’s performance (Meat Loaf). 512 Archives of Acoustics – Volume 43, Number 3, 2018

The highest value of the standard deviation was ob- and the third one) have similar results, while they differ tained for Spectral flatness. Relationships between dif- from the results of the first sample. The largest values ferent performances of the same song can also be found of the standard deviation were obtained for Spectral within the collection analysed. There are songs, for centroid and Rolloff. which regardless who the performer is, the results are There can also be found songs where the param- similar (Fig. 11), such as Blitzkrieg Bop. However, dif- eters depend on which fragment of the song is used ferences may be observed between the original singer for the analysis, and songs where the parameters do and the covers. not depend on the fragment, they are constant for the There are also songs where each performance re- whole track. An example of the first group of songs turns different to some extent results (Fig. 12), such as is a Green Day’s song Minority (see Fig. 14), where for example Love Hurts. The largest differences may be parameter values are similar at the beginning and at observed for zero-crossing, spectral centroid, skewness the end, and they differ from the values of the param- and kurtosis. eters extracted for the middle fragment of the song. Another group includes performances of covers This is especially visible for Zero-crossing and Rolloff. which are similar, but they differ from the original per- The opposite example is Meat Loaf’s song What About former, such as My Heart Will Go On by Celine Dion Love from 2006 (see Fig. 15). Parameter values for (Fig. 13). The parameters of music excerpts contained different fragments of the song are very close to each in Fig. 13 show, that two of the samples (the second other.

Fig. 11. Similar results between diﬀerent performances of the same song Blitzkrieg BOP – song by Ramones (punk rock); covers by: All Time Low (pop punk), Die Toten Hosen (punk rock), Green Day (punk rock), The Clash (punk rock), The Oﬀspring (punk rock).

Fig. 12. Performance results of the song: Love Hurts – song by The Everly Brothers (rock’n’roll); covers by: Cher (pop), Joan Jett (punk rock), Journey (rock). A. Dorochowicz, B. Kostek – A Study on of Music Features Derived from Audio Recordings Examples. . . 513

Fig. 13. Original performance diﬀerent from covers: (My Heart Will Go On – song by: Celine Dion (pop); covers by: Me First and The Gimme Gimmes (punk rock), Sarah Brightman (musical singer).

Fig. 14. Green Day’s song Minority (1 – beginning, 2 – mid part of the song, 3 – end).

Fig. 15. Meat Loaf’s song What About Love from 2006 (1 – Meat Loaf 1, 2 – Meat Loaf 2).

Finally, two interesting cases were analysed. In is very punk-like. It is fast, garage, simply. On the Fig. 16 parameter values for two versions of the same other hand, the newer one (2009) is of a form of a bal- song are shown. The old version of Liebeslied (1988) lad, more rock than punk. However, the instruments 514 Archives of Acoustics – Volume 43, Number 3, 2018

Fig. 16. No changes over time: Die Toten Hosen – Liebeslied – punk rock.

Fig. 17. Change of style: The Calling/Alex Band – Wherever You Will Go – post grunge/indie.

used are the same: vocal, two guitars, bass guitar and shows that it is not obvious whether parameters are drums. The sound is different, but parameter values similar for an artist or for a song, even though they are similar. represent the same style or music genre. Contrarily, Two versions of Wherever You Will Go (see in some cases, the content of the songs is so distinc- Fig. 17), first recorded by the name of The Calling tive, that a performance is not a significant factor. It in 2001, then as Alex Band in 2010, sound similar, is also possible to discern artists who developed their but parameter values differ much. These two versions own characteristic features, which are visible in the are performed by the same band with the same instru- song parametric representation. On the other hand, ments, it is a similar style. There are, however, some there are also artists, whose style evolved over time or differences, the older version is more indie-like, also, they created just one album different from others in when listening to the newer version, one can notice their repertoire. that musicians have more experience, but at the same This quantitative analysis shows that there are am- time some health issues that the vocalist is struggling bitious artists who use various techniques and instru- with. ments. For them, the analytical results are more diverse, and not so unified, but at the same time unique. 4. Conclusions Thus, such a performance may be difficult to be fol- lowed or imitated by others. At the same time, param- The aim of the study was to compare values of eter extracted from songs performed by other more parameters extracted from one song of many perfor- versatile artists are not always that diverse. In some mances, representing a number of music genres, and cases, they are similar to the extend as if they were various songs performed by one musician. The study obtained from one song. A. Dorochowicz, B. Kostek – A Study on of Music Features Derived from Audio Recordings Examples. . . 515

References 14. Kostek B. (2005), Perception-based data processing in acoustics. Applications to music information retrieval 1. Alternative Press, 01.12.2016, and psychophysiology of hearing, Series on Cognitive http://www.altpress.com/index.php/news/entry/what Technologies, Springer Verlag, Berlin, Heidelberg, New is punk this new infographic can tell you York. 2. Barthet M., Fazekas G., Allik A., Thalmann F.B., 15. Kostek B., Kaczmarek A. (2013), Music recommen- Sandler M. (2016), From interactive to adaptive dation based on multidimensional description and sim- mood-based music listening experiences in social or per- ilarity measures, Fundamenta Informaticae, 127, 1–4, sonal contexts, Journal of the Audio Engineering Soci- 325–340, http://dx.doi.org/10.3233/FI-2013-912. ety, 64, 9, 673–682, doi.org/10.17743/jaes.2016.0042. 16. Kostek B., Kupryjanow A., Zwan P., Jiang W., 3. Benward B., Saker M. (2003), Music in theory and Ras Z., Wojnarski M., Swietlicka J. (2011), Re- practice, Vol. I, p. 12, 7th ed., McGraw-Hill. port of the ISMIS 2011 contest: music information retrieval, foundations of intelligent systems, ISMIS 2011, 4. Bergstra J., Casagrande N., Erhan D., Eck D., LNAI 6804, pp. 715–724, M. Kryszkiewicz et al. [Eds.], Kegl B. (2006), Aggregate features and AdaBoost for Springer Verlag, Berlin, Heidelberg. music classification, Machine Learning, 65, 2/3, 473– 17. Kotsakis R., Kalliris G., Dimoulas C. (2012), In- 484, http://dx.doi.org/10.1007/s10994-006-9019-7. vestigation of broadcast-audio semantic analysis sce- 5. Bhalke D.G., Rajesh B., Bormane D.S. (2017), narios employing radio-programme-adaptive pattern Automatic genre classification using fractional fourier classification, Speech Communication, 54, 6, 743–762. transform based mel frequency cepstral coefficient and 18. Lane D.M. (2017), Difference between two means timbral features, Archives of Acoustics, 42, 2, 213–222, (Independent Groups), http://onlinestatbook.com/2/ doi: 10.1515/aoa-2017-002. tests of means/difference means.html 6. Burred J.J., Lerch A. (2014), Hierarchical approach 19. Matlab MIRtoolbox 1.6.2 Specification. to automatic musical genre classification, Journal of 20. Ntalampiras S. (2013), A novel holistic mode- the Audio Engineering Society, 52, 7/8, 724–739. ling approach for generalized sound recognition, 7. Daniels M. (2016), Crowdsourcing the defini- IEEE Signal Processing Letters, 20, 2, 185–188, tion of “Punk”, http://poly-graph.co/punk (retrieved http://dx.doi.org/10.1109/LSP.2013.2237902. 01.10.2016). 21. Palisca C.V. (1998), Marc Scacchi’s defense of new music (1649) [in Polish: Marca Scacchiego obrona 8. Dorochowicz A., Hoffmann H., Majdańczuk A., nowej muzyki (1649)], Muzyka, XLIII, 2, 131–132. Kostek B. (2017), Classification of musical genres by means of listening tests and decision algorithms, ISMIS 22. Pascall R. (2001), The new Grove dictionary of music ‘2017, Springer Series Studies in Big Data, Intelligent and musicians, S. Sadie, J. Tyrrell [Eds.], 24, 2/Lon- Methods and Big Data in Industrial Applications, Be- don, pp. 638–642. mbenik R., Skonieczny L., Protaziuk G., Kryszkiewicz 23. Plewa M., Kostek B. (2015), Music mood visualiza- M., Rybinski H. [Eds.]. tion using self-organizing maps, Archives of Acoustics, 40, 4, 513–525, doi: 10.1515/aoa-2015-0051. 9. Dubnov S. (2004), Generalization of Spectral Flat- ness Measure for Non-Gaussian Linear Processes, 24. Pluta M., Spalek L.J., Delekta R.J. (2017), An IEEE Signal Processing Letters, 11, 8, 698–701, doi: automatic synthesis of musical phrases from multi- 10.1109/LSP.2004.831663. pitch samples, Archives of Acoustics, 42, 2, 235–247, doi: 10.1515/aoa-2017-0026. 10. Helman Z. (2016), The concept of style and mu- 25. Reljin N., Pokrajac D. (2017), Music perform- sic of the twentieth century [in Polish: Pojęcie stylu ers classification by using multifractal features: a case a muzyka XX wieku], Institute of Musicology, Univer- study, Archives of Acoustics, 42, 2, 223–233, doi: sity of Warsaw, Warszawa, http://ksiegarnia.iknt.pl/ 10.1515/aoa-2017-0025. uploads/files/PRM 2006 fragment.pdf. 26. RockSound (2016), http://www.rocksound.tv/news/ 11. Hoffmann P., Kostek B. (2015), Bass enhance- read/study-green-day-blink-182-are-punk-my- ment settings in portable devices based on music genre chemical-romance-are-emo (retrieved 01.13.2016). recognition, Journal of the Audio Engineering So- 27. Rosner A., Kostek B. (2018), Automatic music ciety, 63, 12, 980–989, http://dx.doi.org/10.17743/ genre classification based on musical instrument track jaes.2015.0087. separation, Journal of Intelligent Information Systems, 12. Holzapfel A., Stylianou Y. (2008), Musical 50, 363–384 doi: 10.1007/s10844-017-0464-5. genre classification using nonnegative matrix factoriza- 28. Rosner A., Schuller B., Kostek B. (2014), Classi- tion-based features, IEEE Transactions on Audio, fication of music genres based on music separation into Speech, and Language Processing, 16, 2, 424–434, harmonic and drum components, Archives of Acous- http://dx.doi.org/10.1109/TASL.2007.909434. tics, 39, 4, 629–638, doi: 10.2478/aoa-2014-0068. 13. Kalliris G.M., Dimoulas C.A., Uhle C. (2016), 29. Schedl M., Gómez E., Urba J. (2014), Music infor- Guest Editors’ note. Special issue on intelligent audio mation retrieval: recent developments and applications, processing, semantics, and interaction, Journal of the Foundations and Trendsr in Information Retrieval, 8, Audio Engineering Society, 64, 7/8, 464–465. 2–3, 127–261, doi: 10.1561/1500000042. 516 Archives of Acoustics – Volume 43, Number 3, 2018

30. Seidel W., Leisinger U. (1998), Music in History 33. Soper D.S. (2017), Student t-value calculator [Soft- and the Present [in German: Die Musik in Geschichte ware], http://www.danielsoper.com/statcalc/calcula- und Gegenwart], article: Stil, [Ed.] L. Finscher, pp. tor.aspx?id=10 (retrieval 02.18.2017). 1740–1759, Bärenreiter and Metzler, Kassel. 34. Tekman H.G., Hortacsu N. (2002), Aspects of stylis- 31. Silla C.N., Kaestner C.A., Koerich A.L. (2007), tic knowledge: what are different styles like and why do Automatic music genre classification using ensemble of we listen to them?, Psychology of Music, 30, 1, 28–47. classifiers, IEEE International Conference on Systems, Man and Cybernetics, pp. 1687–1692. 35. Tzanetakis G., Essl G., Cook P. (2002), Automatic 32. Small Music Encyclopedia (1968), [in Polish: Mała En- musical genre classification of audio signals, IEEE cyklopedia Muzyki], Śledziński S. [Ed.], PWN, War- Transactions on Speech and Audio Processing, 10, 5, szawa. 293–302.